Developing essential fish habitat maps: report

4. Discussion

This study developed the methodological framework and maps to identify potential essential fish habitats for 29 fish and shellfish species that are relevant to Scottish waters for their commercial or ecological importance (including 13 species that are designated as Priority Marine Features in Scotland's seas; NatureScot 2020). Where fish survey and environmental data were available, species distribution models were applied to map the potential EFH based on the distribution of aggregations of the species as a whole (for habitats used as refuge) or of their specific life stages (using juveniles to indicate nursery habitats, spawning adults or eggs for spawning grounds). For inshore areas, where suitable data were lacking, a habitat proxy approach was applied instead. This relied on literature review and expert input to quantify the importance of different inshore EUNIS habitat types in supporting the species during their life cycle, or at particular stages of it (e.g. juveniles, spawning adults, eggs).

Distribution mapping for several of the species considered in this study has been undertaken before, particularly for gadoids and flatfish of commercial interest and some elasmobranchs (e.g. Coull et al. 1998, Ellis et al. 2012, Aires et al. 2014, González-Irusta and Wright 2016a, 2016b, 2017, Franco et al. MMO 2013, 2016, AFBI 2021, Katara et al. 2021, Langton et al. 2021), and, recently, also for brown crab (Mesquita et al. 2021). The most recent of these studies adopted species distribution modelling to infer the spatial distribution of life stages of these species at a high spatial resolution and over decades based on environmental predictors. However, several of these applications did not cover Scottish waters (Franco et al. MMO 2013, 2016, AFBI 2021, Katara et al. 2021), or focused only on juveniles (Aires et al. 2014). In turn, less recent studies (Coull et al. 1998, Ellis et al. 2012) derived UK-wide maps directly plotting survey catch data, resulting in a snapshot of the observed distribution in a restricted period of time (e.g. in 2010 in Ellis et al. 2012) and at a low spatial resolution (ICES rectangle). The present study has expanded the high- resolution distribution mapping of these species into Scottish waters, and indeed with UK- wide coverage for several species. A high spatial resolution is a pre-requisite for such spatial products to be used as supporting tools for marine spatial planning hence to make decisions on siting of activities at sea (e.g. offshore wind farms). Furthermore, to our knowledge, this study is the first to provide habitat mapping for some shellfish species that were poorly covered in previous studies (e.g. long finned squid, European lobster, velvet crab, cockles, clams and whelks).

4.1 Weight of evidence assessment

Aires et al. (2014) suggested that the maps resulting from species distribution model predictions should be used as an additional tool to complement existing information, rather than replacing it. These authors also called for stakeholder input (particularly from the fishing industry) to complement the outputs. Both these aspects were addressed in the present study by applying a 'Weight of evidence assessment' approach and undertaking map validation through consultation of stakeholders, including the fishing industry.

The 'Weight of evidence assessment' approach (EFSA 2017) was used in this study to identify and locate habitats with particular functional roles (as refugia, nursery or spawning) for the studied species in the marine environment. This approach relies on integrating multiple lines of evidence from different sources to answer the question about the location of such habitats. Specifically, the maps developed and presented in this report integrated (i) evidence on the distribution of the species as directly observed in fish surveys, (ii) the potential distribution of their habitats as predicted through species distribution models (based on survey and environmental data) or habitat proxies (based on literature and expert knowledge), and (iii) expert assessment based on stakeholder validation.

Any individual line of evidence had its strengths and weaknesses:

• i. Survey data alone provided an accurate representation of the actual distribution of a species at a certain point in time in most of the cases, but had limitations in their spatial and temporal coverage, or in their ability to represent certain species and/or life stages (depending on the survey method affecting catchability; e.g. Nephrops). For example, the International Bottom Trawl Survey data used in this study were widely distributed in UK waters, but with poor coverage of inshore waters, and poorer or no adequate representation of certain species (e.g. pelagic species, benthic shellfish). In turn, other surveys that extended further inshore were often restricted to smaller scale areas (Sandeel Dredge surveys, West Coast of Scotland Demersal Fish (WCDF) survey) and/or provided information on the species distribution for one year only (WCdf 2013/14).
• ii. Predictions obtained from species distribution models or using habitat proxies allowed to extrapolate the distribution of the species even in areas where no direct observations are available, thus providing indication of areas that may be potentially used by a species or its life stages. This was based on suitability of the specific environmental or habitat characteristics considered in these assessments, but did not account for the species interactions with other factors that might also affect the actual use of the area, for example, natural biotic interactions such as competition or disease. These are seldom included in predictive models due to the difficulty in measuring these complex relationships, and may result in the overprediction of the species distribution (Velazco et al. 2020). Efforts to account for biotic factors were made in this study by including net primary productivity as a potential indirect predictor for food availability in the EFH models. However, other direct interactions with prey, predators or competitors which might also affect the distribution of a species could not be included, nor anthropogenic factors, which may also alter the natural environmental/habitat suitability of certain areas (e.g. through impacts on the seabed). Therefore, predictions obtained with these assessments were based on a simplified representation of the interactions of a species with the marine

environment, with consequent limitations in the ability to accurately represent the species distribution. The confidence assessment undertaken for the EFH models and habitat proxies in this study provided an indication of the extent of such limitations, while also accounting for limitations in the evidence (data, literature, expert knowledge) supporting these predictions.

• iii. Expert input from stakeholders provided a broader assessment of the species habitat distributions represented in the maps, but this line of evidence is also not exempt from potential biases, e.g. due to the specific expertise and experience of the individuals. Consultation with a high number of stakeholders with highly diverse expertise (e.g. on different species) would allow to reduce the effect of these biases, although this requires time (also depending on stakeholder availability), which was limited in a short-term study such as this one.

The spatial outputs provided in this report should be read considering the combination of all these lines of evidence to indicate the distribution of the species and their possible EFH. The confidence shown on the map is related to the model or habitat proxy assessment that led to the map predictions (line of evidence ii above). This confidence should be used as a baseline to guide consideration about the relative validity of the predicted results for different species (with more weight to be given to outputs with higher confidence, i.e. closer to 100%, as shown in Table 29 and Table 30). However, a higher overall confidence is associated with the final spatial product where the multiple lines of evidence were combined. In fact, the 'Weight of evidence assessment' approach allows for individual lines of evidence to partly compensate for each other's failings (e.g. the comparison with additional survey data and/or the expert input obtained from stakeholders during map validation allowed to identify areas where the EFH is known to occur but that were 'missed' by the EFH model prediction; see Table 31). This balances the biases of the individual lines of evidence considered, thus improving the overall confidence in the integrated output. The confidence is reinforced particularly where different lines of evidence converge towards the same result and no discrepancies were observed (EFSA 2017).

4.2 Operationalising the 'Essential fish habitat' concept

Essential fish habitats (EFH) are defined based on the function they perform for a species (spawning, nursery, etc.; U.S. Magnuson-Stevens Fishery Conservation and Management Act 1976). Such function is often associated with a specific life stage or phase of the life cycle of the species (e.g. juveniles for nursery), so that an EFH would be part of the area where that life stage occurs (as per conceptual representation in Figure 2). Whether a habitat for a given life stage can be identified as an EFH depends on whether the habitat is able to improve the condition of that life stage and thus giving an advantage (or added value) that other habitats are not able to provide (or provide to a lesser degree), with a resulting benefit for the population as a whole. For example, Beck et al. (2001) identified this advantage for nursery habitats as a greater contribution per unit area to the production of individuals that recruit into adult populations compared to other habitats where the juveniles were present (see difference between habitat B and C in Figure 2). This greater contribution can be realized through increased density of individuals in the habitat, improved or faster growth of the individuals and of the population (i.e. greater survival and lower mortality), increased biomass production, and increase of the rate of successful export of this biomass into the wider population.

The combination of all the factors mentioned above likely results in an 'added value' associated with an EFH compared to other habitats used by the species/life stage. However, data measuring some of these factors are not always easily obtained, and occurrence and abundance data available from wide-scaled, repeated monitoring programmes are used instead. This approach was adopted for the identification of habitats potentially providing key EFH functions through species distribution modelling in this study, whereby aggregations of juveniles, adults in spawning conditions or eggs were used as indicators of potential nursery or spawning grounds, respectively. For EFH functions that benefit multiple life stages of the population (e.g. refuge function or mixed functions provided by a habitat), aggregations of individuals irrespective of their life stage were considered. The assessment of aggregations (as opposed to considering the mere occurrence of a species or life stage) was assumed to provide an indication of habitat with potential added value for their ability to support higher densities, hence reflecting one of the elements characterizing EFH (sensu Beck et al. 2001).

Specific criteria (e.g. body size, seasonality, spawning condition) were applied to identify the life stages of interest, and therefore their aggregations for species distribution modelling, and the results (EFH models and maps) are therefore dependent on these choices.

Standardisation of these criteria with previous assessments was undertaken, where possible, so that consistency and comparability was allowed. For example, 0-group individuals were identified as indicators of juvenile habitats potentially functioning as nursery for most species (e.g. as in Ellis et al. 2012, Aires et al. 2014), although in some cases 1-group fish were also included where the survey data for 0-group alone were insufficient.

Aires et al. (2014) suggested this approach for saithe, for example, on the account that juveniles of this species use inshore nursery habitats until they are 2 - 3 years of age, thus allowing to expand the age group considered while still representing nursery areas.

Similarly, common sole juveniles are known to spend 2 - 3 years in inshore nursery grounds before migrating offshore (Rijnsdorp et al. 1991, ICES 2012), and therefore 1-group individuals were also considered in this case. The size criterion used for anglerfish juveniles also led to the inclusion of some 1-group individuals in the analysis to compensate for the scarcity of catch data available for 0-group individuals only. This was done on the account that the species does not reach sexual maturity until 5 years of age at least (Laurenson et al. 2001).

Where the habitat proxy approach was applied for inshore waters, it would be incorrect to identify its results as EFH. In fact, this assessment only accounted for the association of the life stage (relevant to an EFH function) with the habitat as reported in the literature. The supporting evidence was often based on observations of occurrence, resulting in an indication of key habitat for the species (habitat B in Figure 2) rather than on abundance (hence aggregation) or other more detailed parameters (increased growth etc), which would have allowed an indication of EFH (habitat C in Figure 2), as it was done in the data-based EFH modelling approach. Therefore, the habitat proxies identified through this method may overestimate the distribution of a species (across the inshore habitats assessed at least), and its EFH is probably a subset of the mapped potential distribution (as outlined in Figure 2).

4.3 Important Data/Knowledge Gaps identified in the EFH Assessment

The collation of data and evidence in this study highlighted specific areas where gaps in the evidence available (data or knowledge) occurred and that need further work. A summary of the identified knowledge and data gaps is given in Table 30 and Table 31, respectively. It is highlighted that the relatively short time available to develop this project was a limiting factor to the extent of the literature review, data identification, collation and analysis that could be undertaken. This may have influenced the identification of knowledge/data gaps, and is taken into consideration in the discussion below, where additional evidence was known to exist but it could not be obtained within this project.

Table 30. Summary of knowledge gaps identified from literature review in this project and relevant to habitat proxy assessment for fish and shellfish species/life stages inshore.

Species	Knowledge gap
Sandy ray; common skate	No knowledge about habitat preferences of juvenile/spawning. Habitat proxy assessment could not be undertaken.
Thornback ray	Little knowledge of specific environmental preferences for egg- nurseries. Reduced confidence in habitat proxy assessment.
Sprat; common sole; spotted ray	Scarce knowledge and little detail about juvenile habitat preferences. Reduced confidence in habitat proxy assessment.
Spurdog	Scarce knowledge and little detail about parturition habitat preferences. Reduced confidence in habitat proxy assessment.
Saithe; cod; whiting; European lobster; brown crab; velvet crab	Little detail about juvenile habitat preferences (often limited to depth). Reduced confidence in habitat proxy assessment.

Table 31. Summary of data gaps identified in this project.

Feature	Data gap in project	Possible additional data sources
Shellfish, offshore: queen and king scallop; surf clam (+ possibly dog cockle, common whelk)	No offshore shellfish survey data obtained. EFH models could not be calibrated.	MS scallop dredge surveys (Scotland)(1). Other additional data sources not known(2). Possible need for further monitoring for species/geographic areas not represented by dredge survey data.
Shellfish, inshore: European lobster (juvenile); brown crab (juvenile); velvet crab; razor clam; common and dog cockle; common and dog whelk.	No inshore shellfish survey data obtained. Habitat proxy maps could not be validated. EFH models could not be calibrated inshore (see also data gaps for environmental layers inshore)	UHI Shellfish Stock Assessment Programme (Shetland, since 2000; relevant to crabs and common whelk). SAMS/MS razor clam surveys (electrofishing with towed video; Firth of Clyde). Offshore wind industry crustacean monitoring for impact assessment (accessibility to be ascertained on a case-by-case basis). Other additional data sources not known. Possible need for further monitoring for species/geographic areas not represented by the survey data above.
Fish, inshore: juvenile stages of common sole, plaice, cod, whiting, saithe, Norway pout, anglerfish, sprat, skates, rays, spurdog (parturition grounds) (also juvenile long finned squid)	No inshore survey data obtained. EFH models could not be calibrated inshore (see also data gaps for environmental layers inshore)	WFD fish monitoring programme for transitional and coastal waters (since 2000, at least; EA data for England and Wales readily available from online database; accessibility of SEPA data for Scotland to be checked). Other additional data sources not known, but possibly from multiple local research/ inshore monitoring studies. Possible need for further monitoring for species/geographic areas not represented by the survey data above.
Environmental and EUNIS habitat data layers, inshore	Reduced spatial coverage in more inshore areas (e.g. sea lochs, areas closer to shore) by environmental and habitat data layers. Resulting in spatial gaps in habitat proxy maps, and limiting ability of calibrating EFH models from inshore survey data.	Additional data sources not known, but possibly from multiple local inshore environmental/habitat monitoring studies. Possible need for further environmental/habitat monitoring for geographic areas of interest where gaps persist.

Table footnotes: ⁽¹⁾ Scallop dredge survey data may also provide suitable data on juvenile stages of anglerfish and thornback ray, to integrate/calibrate EFH models for these species/life stages. ⁽²⁾ The use of VMS data was suggested to identify boundaries of fishing grounds, but these data would not provide sufficient information (e.g. on species CPUE to identify aggregations) or at the required resolution (higher than the broad scale area) needed for the EFH modelling. However, these data could be used for map validation of EFH model map for shellfish, should these be developed in the future.

Lack of or poor ecological knowledge on specific habitat preferences existed in the literature for some species (Table 30). This gap was particularly relevant for skates and sandy ray as it did not allow the assessment of habitat proxies for their inshore habitats, whereas it led to a reduction of the overall confidence in the assessment of other species (Table 32). This is a notable knowledge gap, also considering that most of these species are of commercial relevance as food products (e.g. lobster, crabs, cod, whiting) or of conservation importance (e.g. endangered/critical state of ray and skates as per IUCN red list). As mentioned above, the literature review was by no means exhaustive, but sufficient evidence could be obtained for several other species, suggesting this reflects a real knowledge gap that should be addressed through further targeted research so that more can be learned about these species' habitats (with particular regard for the life stages mentioned in Table 32).

Furthering knowledge on these species would make it easier to predict their key habitats with higher confidence and, if needed, this would then enhance efficacy of protection efforts and marine spatial planning.

There was also a clear gap in data availability for shellfish, with the exception of Norway lobster and the squid species considered in this study (Table 33). No survey data could be obtained for European lobster, brown crab, velvet crab, queen and king scallops, common and dog cockles, surf and razor clams, and common and dog whelks. This led to the inability to calibrate data-based EFH models for these shellfish species, or to validate the habitat proxy maps developed for inshore habitats used by some of these species (Table 33). It is of note that several of the shellfish species considered here have their main habitats inshore, in some cases restricted to the intertidal fringe (e.g. common cockle and dog whelk).

Therefore, existing surveys for these species are likely to be small scale (compared to the wider fish survey offshore) and/or short term.

Some data sources were identified in the project, but the data could not be obtained within the available timescale. For example, Marine Scotland Science regularly undertakes scallop dredge surveys that may provide suitable data for the calibration of EFH models for scallops and other shellfish occurring offshore (Table 33). Stakeholders have also indicated the potential use of these surveys to provide data on juvenile anglerfish and juvenile thornback ray. Surveys for razor clam have been recently undertaken along the Ayr coast, in the Firth of Clyde to test a new sampling method (Fox et al. 2019), and a Shellfish Stock Assessment Programme (Shetland UHI 2022) has been undertaken by the North Atlantic Fisheries College, University of the Highlands and Islands since 2000, involving inshore sampling in the Shetland targeting in particular velvet crab, brown crab and common whelk. The private sector is also a possible source for shellfish data. For example, crustacean monitoring is often undertaken by the offshore energy industry to assess impacts on lobster and crab stocks (e.g. Roach et al. 2018). If they can be accessed, these data (particularly from control/unimpacted areas) may be used to develop EFH models for lobster and crab species (particularly where the data include information on juvenile catches). In any case, individual local surveys alone would provide limited evidence on the wider distribution and general habitat preferences of the species, and the combination of evidence from multiple surveys (possibly with cumulative wide geographical distribution) would be needed to calibrate EFH models that have wider applicability. Such data gathering requires more time and effort to identify and source the data, where these can be made available.

Limited data from inshore fish surveys could be obtained for this project. Although they were used for the validation of some of the habitat proxy outputs, they could not enough to calibrate EFH models, due to the limited representativity of these data21 (covering one year of survey on the west coast of Scotland only). The collation of additional inshore fish survey data could allow the EFH modelling of inshore habitats for fish species, and further validation of the habitat proxy mapping around the Scottish (and UK-wide) coastline. A potential useful source of data in this case would be from the fish monitoring of transitional and coastal waters that environment protection agencies around the UK have undertaken since 2000 to comply with the Water Framework Directive. While these data are publicly available from the online database held by the Environment Agency for English and Welsh estuaries (WFD TraC fish count data, Environment Agency 2021), the survey data for the Scottish TraC waters are currently unavailable due to a hacking event on SEPA IT in 2021.

Therefore, these data could not be obtained in this study, and further efforts could be made once the problem is resolved.

Limitations in the spatial coverage of inshore areas by environmental data layers might still reduce the ability of calibrating species distribution models for fish and shellfish species

inshore. Maps for the water quality variables used in the models (e.g. sea temperature, salinity) are obtained from the application of broad scale, oceanographic models which do not provide reliable data for marine areas closer to shore and internal waters (e.g. sea lochs), where land influences occur with higher spatial and temporal variability. Spatial gaps were also observed in the EUNIS habitat data layers used for the habitat proxy assessment, particularly in internal areas (e.g. sea lochs) that may be used as EFH by some species (e.g. cod). Environmental/habitat surveys targeting these most inshore areas where data are lacking could help filling these gaps.

4.4 EFH model outputs

Where the data allowed it, species distribution models were applied for the UK-wide mapping of the potential important habitats for fish and shellfish species, as inferred from the environmental suitability for aggregations of relevant life stages. The nature and distribution of the fish survey data used to calibrate the EFH models affected the spatial coverage of the mapped predictions (as the EFH model predictions were considered to be valid only within the environmental ranges represented in the survey data that were used to calibrate the model) as well as the overall confidence in the resulting model.

Most of the predicted maps for fish and shellfish EFH models are based on bottom trawl and beam trawl data had poor coverage of most inshore areas, partly due also to the lack of coverage of these areas by environmental data layers (e.g. for variables obtained from oceanographic models). This was an issue in particular for the mapping of plaice, common sole, sprat, whiting and long finned squid, for which nursery habitats are known to occur closer to shore. The limited distribution of the collated survey data in inshore nursery areas, hence their limited ability to fully represent these primary nursery habitats, also likely contributed to the relatively low confidence of some of these EFH models (e.g. whiting and squid). For the same reason, a model for juvenile cod could not be calibrated due to insufficient data, and therefore the collation of additional survey data from inshore areas would also benefit the modelling of this species.

In turn, survey data for Ammodytes marinus were restricted to sandeel grounds located on sand and coarse sediments relatively inshore (<66 km from the shore, at depth <100 m) off the east coast of Scotland. The resulting map provided limited spatial coverage, as it did not allow predictions for offshore sandbanks that might function as sandeel EFH (e.g. Dogger Bank and North Norfolk sandbanks; Van der Kooij et al. 2008, Langton et al. 2021).

Expanding the survey database for the sandeel EFH models (e.g. with data from sandeel habitats on the west coast of Scotland and in NC MPAs, on Dogger bank and Norfolk sandbanks) would allow the validity of these model predictions to be extended to wider environmental ranges (over space and, likely, time), thus improving the geographical coverage in the maps. This would also likely increase the overall confidence of the sandeel EFH model predictions, as the spatial (and temporal) coverage of the surveys was one of the elements considered to assess confidence in the survey data.

Including survey data from targeted surveys for individual species was also a way to improve the EFH model overall confidence, as this also accounted for the assessment of the survey data used in the model. In fact, a higher catch efficiency of the survey method is expected from surveys targeting individual species (e.g. sandeel dredge survey, anglerfish surveys), compared to other broader-target surveys (although a trade-off could occur with spatial coverage, as observed for the sandeel surveys), thus contributing to a higher confidence in the survey data. For Nephrops, broad scale bottom trawl surveys were used to predict aggregations of individuals as a proxy for areas with higher density of burrows, hence with an added value as refuge habitats. Although bottom trawling is a common fishing method for the species, the emergence of individuals from the burrows (to feed or mate) may be unpredictable and therefore bottom trawl catch data may not be optimal for an accurate assessment of Nephrops' burrow distribution. Direct observations of burrow density are available from TV video surveys undertaken by Marine Scotland Science (ongoing since 2007). Such data (for the period 2007-2016 only) were obtained for the present study, but could not be included in the EFH modelling (due to timing issues) and were only used for map validation. Inclusion of these surveys in the database for the EFH model (as aggregations of burrows), along with the assessment of aggregations at the stock level (as suggested in stakeholder feedback), would likely lead to better model predictions with higher associated confidence.

The environmental data used to calibrate the EFH models included both persistent and non- persistent (i.e. temporally variant) variables. The latter (obtained from data layers for water temperature, salinity, water column mixing and net primary production) were specifically included to allow the EFH model to capture temporal as well as spatial dynamics in the distribution of the species. The incorporation of non-persistent data is an essential element to allow predictions under temporally variable environmental scenarios, including for example climate change (see Annex 4). Almost all EFH models that were calibrated were dynamic tools that included at least one non-persistent environmental predictor (the only exception was the EFH model whiting juveniles which only included persistent variables as predictors of the aggregations of this species). However, mapped predictions are a static representation of the EFH model applied to a specific environmental scenario, and the choice of the latter may affect the resulting spatial outputs and their accuracy in representing the species distribution. This was evident for the EFH model maps developed in this study, for which the underpinning environmental scenario was based on the average conditions over the studied period 2010 - 2020 (for the season relevant to the specific life stage). As such, the maps were not able to fully represent the temporal variations (between years and even within a season) in species distribution or persistence at given sites.

Some of the inaccuracies highlighted by the stakeholder validation in the mapped EFH model predictions could be ascribed to the average environmental scenario used rather than to a failing of the model predictive ability. This was clear, for example, for maps for lesser sandeel and anglerfish, two of the top-ranking EFH models in terms of model statistical performance and overall model confidence. The comparison of these with maps predicted on more accurate temporal scenarios (for individual years) showed a better correspondence with the data and expert knowledge on map validation. Therefore, this possible limitation in the static map (but not in the dynamic model) should be taken into consideration when using the maps. As this does not affect the validity of the EFH model themselves, but rather of the mapped predictions, the EFH models can be applied to a specific scenario of interest (e.g. to ascertain potential distribution of the EFH in a certain year or under specific environmental conditions) to obtain a more accurate predicted map (see Annex 2 for guidance on how this can be easily done, without the need for specific statistical knowledge or tools other than the EFH model decision tree diagram). In turn, the more general maps shown in this report could be improved to better capture the temporal variability and site persistence in the species distribution by applying the EFH model to multiple temporal environmental scenarios (e.g. environmental conditions in different years) and then combining the resulting spatial predictions (e.g. by average, frequency of results). This was not possible within the short timescale of this study. A similar approach has been used for example by Katara et al. (2021).

4.5 Habitat proxies

The use of habitat proxies facilitated the development of indicative maps in inshore waters where data was lacking to develop EFH models. Habitat proxies were applied to species for which essential habitats occur mostly inshore. These habitat proxy assessments carried variable confidence, depending on the amount and detail of the supporting evidence and expert knowledge. Higher confidence was generally associated with the assessment of species that had clear and very specific habitat preferences (e.g. sandeel, herring, cockles), as opposed to species for which there was poor knowledge (e.g. elasmobranchs) or for which habitat characterisation was more generic. A more generic habitat characterisation could possibly reflect more generalist habitat preferences, or just the poor detail in the characterisation of the habitat for these species, which often did not allow to discriminate suitability at the higher habitat resolution (EUNIS habitats Level 4) (e.g. cod, whiting).

In the case study area mapped for habitat proxies (west coast of Scotland), available survey data showed that higher abundance of juveniles (mostly 0-group individuals, with 1-group also considered for some species as described for the EFH models) in the catches appeared to often correspond with areas on or in near proximity to areas predicted as potentially suitable for the species. Considering that fish juveniles are mobile, and therefore may also be detected in trawl surveys near their preferred habitat, these results were considered as a fair validation of the habitat maps. However, some species (elasmobranchs and saithe) were poorly represented in the fish survey catches, and, with these data being from surveys undertaken over one year only (2013/14), little information could be obtained about persistence of use of the observed areas. No survey data could be obtained for benthic shellfish (European lobster, brown crab, velvet crab, queen and king scallops, common and dog cockles, surf and razor clams, and common and dog whelks) for the map validation. Gathering further survey data from inshore surveys (for fish and shellfish) in the study area would improve validation of the habitat proxy maps produced here. Also, additional stakeholder consultation of these maps would be beneficial to further validate the habitat maps providing an additional line of evidence.

Similar to the EFH model maps, the use of habitat proxies provided an indication of areas that may be potentially suitable for functional use by the species/life stages, and which likely include EFH for the species (see sections 2.3.1 and 4.2). However, factors other than the habitat type may also influence the actual functional use of the habitat by a species, including for example other abiotic conditions not accounted for in the habitat type classification (e.g. water temperature), biotic interactions or anthropogenic impacts. As a result, it is possible that the suitable functional habitats in the maps may overestimate the extent of habitats used by the species (at least within the range of inshore habitats that were assessed in this study). Improving the map validation via comparison with additional survey data and expert knowledge would confirm if and where habitats of a species may be over predicted. In fact, habitat proxies identify suitable functional habitat areas that may be more extensive than the actual EFH, and this may be indicated by areas where the functional habitat is identified by habitat proxies but for which there is no evidence from survey data on the presence of aggregations of the species/life stage.

As observed for the EFH model maps, the choice of the mapped environmental layers (EUNIS habitat data in this case) to spatially predict the results of the assessment (as per habitat proxy matrix in this case) had also an influence on the spatial output. The assessment of habitat proxies for the fish and shellfish species was focused on inshore habitats, and this led to the a priori exclusion of deep circalittoral habitats (e.g. EUNIS sedimentary habitats A5.15, A5.27, A5.37), which were often identified as "offshore (deep) circalittoral habitats" of habitats "in the offshore circalittoral zone" in the EUNIS habitat descriptions22. On mapping the habitat proxies in the case study area, it became apparent that these EUNIS habitats may occur close to the shore. A choice was made to prioritise mapping at the higher resolution, both spatial and in terms of habitat classification (EUNIS Level 4), so to provide a more accurate spatial representation of the habitat proxy assessment as undertaken via the matrix tool. However, this resulted in spatial gaps in the maps for the non-assessed habitats (along with assessed habitats that could not be scored due to lack of supporting evidence of use). The habitat proxy maps could have been drawn by only considering Level 3 habitats (including all the possible sub-habitats within) and their respective scoring in the assessment matrix. However, this would have likely led to an overestimation of the spatial distribution and suitability of the habitats for a species. This could have resulted in the score allocated to the EUNIS Level 3 habitat also attributed to areas in the map where, at the finer resolution (Level 4), parent habitats that were scored lower, were not scored (due to lack of evidence) or were not included in the assessment (deeper habitats) occurred. This was not considered suitable for the purpose of the maps in supporting marine spatial planning, and therefore, a higher accuracy and resolution of the spatial products was favoured, despite the gaps. These gaps could be filled by integrating the assessment with additional consideration of the excluded deeper habitats. Whether a score could be allocated to these additional habitats would depend on the re-examination of the available literature to determine whether there is evidence of their use or on expert input from stakeholders.

Within the time allowed in the project, mapping of habitat proxies was only undertaken for the case study area on the west coast of Scotland as an example of the spatial implementation of the assessment. However, the assessment undertaken in this study (as per habitat proxy matrix) has a wider validity and was spatially applied to EUNIS data layers available at the appropriate habitat level (Levels 3 and 4) in UK waters. Therefore, the spatial implementation of the assessment can be extended to the whole of inshore Scottish waters (or indeed to the whole of the UK coastline, if desired) for a more comprehensive mapping. However, considering the small scale of the suitable habitat areas for certain species, a combination of finer scale maps for different case study areas as opposed to a single map for the whole extent of Scottish inshore waters would be preferable to better represent the results of the habitat proxy assessment.

Whichever spatial implementation is applied, the confidence associated with the habitat proxy assessment (matrix) should be always considered as an integral part of the assessment to weight the results being considered for a species. However, for the spatial implementation of the habitat proxies, the variable confidence in the EUNIS habitat data used as base layers should also be considered. In fact, EUNIS habitat data layers are available at different spatial resolution, with a trade-off existing between this and the spatial coverage and confidence. More resolved EUNIS layers are available from habitat surveys (from different years and locations), thus providing a direct assessment of the habitat distribution with higher confidence, albeit often covering smaller areas. In turn, broadscale EUNIS layers are interpolated from models, hence carrying a lower confidence compared to direct observations, but allowing coverage of wider areas. Both types of data layers were used for habitat proxy mapping, although, where available, data from most recent surveys were prioritised over broadscale habitats predicted for the same locations. This was considered a good compromise between resolution, confidence and coverage of the resulting spatial implementation. Where a different selection of the base EUNIS habitat maps is made (e.g. use of broadscale data layers only to increase spatial coverage), the loss in spatial resolution and confidence should be considered.

4.6 Potential implications of climate change

Different species may respond differently to environmental change, depending on their environmental tolerance, preferences and optima. Changes in the abiotic conditions of the marine environment may also affect species indirectly, e.g. by influencing other biota with which the species interact (e.g. through prey-predator, competitor, host-parasites relationships) or biota which provide other benefits to the species (e.g. habitat structuring species such as aquatic vegetation or biogenic reef). The variability in the species response was confirmed by the predictions for their potential distribution under environmental change (namely sea bottom temperature and/or depth), where major changes were predicted for lesser sandeel and anglerfish, whereas the changes in the distribution of Nephrops and juvenile plaice were marginal. In addition, while the changes for juvenile anglerfish suggested a redistribution of areas potentially suitable as nursery grounds towards more inshore waters and the northern North Sea, a notable loss of potentially suitable areas for sandeel was predicted under the scenario of environmental change.

It is emphasised that the observed predictions tested in this document under the environmental change scenario should by no means be considered as accurate projections of the distribution of the species' suitable habitats in the next 100 years under climate change effects. Rather, they should be read as an indication of the potential sensitivity of the relative distribution of essential fish habitats to some environmental changes of similar magnitude and direction as those predicted with climate change conditions. There are different reasons for this. First, the changes applied to sea bottom temperature and water depth inshore (following sea level rise) were derived starting from a different temporal baseline (year 2015) than in the UKCP09 projections (years 1975 and 1985, respectively). In addition, the scenarios of environmental change applied here only account for change in either sea bottom temperature and/or water depth, while there are other environmental variables that might vary with climate change and that are important predictors in the EFH models considered.

Salinity, for example, is one factor that contributed in predicting the distribution of all the species/ life stages considered in this annex, as well as of several other species that were modelled in the study (e.g. juvenile sprat and common sole, spawning cod and whiting, mackerel eggs; see main report). In addition, the assessment of habitat proxies for the distribution of species /life stages inshore as applied in this study often identified a variety of habitat types at reduced or variable salinity as potentially highly suitable, for example for the small sandeel A. tobianus, spawning herring, and juvenile common sole, cod, whiting and sprat (see results for these species in section 3). While the marine projections for salinity under climate change suggest a minor decrease (by 0.2 salinity units) across the entire northeast Atlantic and the North Sea, it is acknowledged that they mainly reflect wide scale changes in the ocean rather than the local effect of rivers (Marine Scotland 2017a, Strong et al. 2017). At the local scale, inshore, the latter effect on salinity might become predominant (e.g. in estuaries and firths). Projected climate changes for the 21st century predict wetter winters, with a 25% increase in river flows, and drier summers, with a 40-80% decrease in the mean flow particularly in upland western regions of the UK (Strong et al. 2017). This has the potential to lead to alteration of the balance between marine and fresh waters in estuarine systems. Sea level rise may also contribute to affecting saline intrusion

in estuaries: a SLR of 1 m has been predicted to increase saline intrusion length by >7 % in deep estuaries and >25 % in estuaries shallower than 10 m (Prandle and Lane 2015). Past climate-induced shifts in estuarine hydrological and salinity conditions were identified as the main factor responsible for changes in the abundance of fish species using some estuarine systems as nursery ground (Pasquaud et al. 2012, Chaalali et al. 2013). For marine species such as those mentioned above, able to use habitats at reduced or variable salinity, a climate-induced variability in the saline intrusion into the estuaries might lead to marked changes in the extent of estuarine habitats suitable for use by these species, also depending on the seasonality of the life stages using these habitats (e.g. juvenile fish use often peaking in the spring-summer). For example, an increased use of the Gironde Estuary (France) by sprat was observed during past events of increased intrusion of marine waters as a consequence of a decreased river discharge and river runoff, while opposite trends were observed for the estuarine flatfish flounder (Platichthys flesus) and catadromous species such as smelt (Osmerus eperlanus) (Pronier and Rochard 1998, Pasquaud et al. 2012, Chaalali et al. 2013).

Climate-induced changes may also lead to spatially variable alterations of the water column mixing, hydrographic circulation and wave height (likely to affect current and wave energy and their distribution) (Strong et al. 2017). These were often important predictors for the species/life stages considered here and for others of the species modelled in this study. The energy of the environment (as accounted by waves and currents) is also one of the factors determining the EUNIS habitat classification, and therefore changes in current and wave energy conditions might affect the habitat distribution in inshore areas, hence their use y fish and shellfish. For example, predicted decreasing trends in wave height (–0.3 cm yr-1) north of Scotland (Strong et al. 2017) might reduce, in the long term, the energy of high- moderate energy habitats that have been identified as highly suitable for example for use by juvenile saithe, dog whelk (see results of habitat proxy assessment for these species in section 3), thus potentially altering their suitability for these species.

In turn, on the intertidal zone, sea level rise is likely to enhance wave and tidal energy (due to increased water depth), thus likely affecting sediment processes (erosion, deposition, transport) (Strong et al. 2017). Through sea level rise, climate change has the potential to modify the estuarine and coastal topography thus further altering the availability, configuration, and location of habitats (Taylor et al. 2004, Fujii and Raffaelli 2008, Masselink and Russell 2013). Effects are expected to be particularly marked where the habitat high water mark is residing against a hard defence structure such as a sea wall, which would prevent the shore to naturally retreat, leading, eventually, to the loss of the intertidal sediment through 'coastal squeeze' (Pontee 2013). Similarly, the area cover of intertidal rocky shores is likely to decline with sea level rise (Robins et al. 2016). While the EFH models in this study provide only partial coverage of inshore waters, the habitat proxy assessment identified the importance of shallow nearshore habitats for several fish and shellfish species, with intertidal (littoral) habitats being often indicated as highly or moderatelysuitable, for example, for juveniles of saithe, common sole, plaice, cod and whiting, and for common cockle and dog whelk (see results of habitat proxy assessment for these species in section 3). The loss of intertidal habitats (including tidal flats and saltmarshes) resulting from cumulative impacts of a range of anthropogenic pressures is already considered an extreme pressure in estuarine areas (Colclough et al. 2010), and has been suggested to reduce the attractiveness of these systems for fish use (Amorim et al. 2017). Climate-induced changes might locally exacerbate such loss of intertidal habitats, reducing their availability to fish and shellfish, with possible bottleneck effects on the size and productivity populations (e.g. where habitat loss reduces availability of estuarine nursery habitats, Pörtner and Peck 2010).

While the potential effects on the availability and distribution of important habitats for fish and shellfish has been discussed above considering some of the environmental factors of change individually, these may interact with each other and with other abiotic and biotic variability, as well as with local anthropogenic pressures, affecting the habitat use by fish and shellfish. For example, effects on wind intensity and patterns, and ocean circulations will have the potential to affect the connectivity between estuarine and marine habitats, through effects on the larval transport from spawning to nursery grounds. Also, increased residence times in transitional waters would likely reduce the dilution of dissolved nutrients and pollutants and increase the time to flush then from the system (Struyf et al. 2004), thus possibly affecting the habitat quality (e.g. through increased risk and frequency of algal blooms and consequent low oxygen conditions, greater exposure to pollutants; Graham and Harrod 2009), hence its suitability and use.

4.7 Conclusions and recommendations

The recommendations about using and improving the assessment tools and associated maps developed in this report are outlined in Table 32, with additional details and conclusions provided summarised in the text below

Table 32. Final recommendations.

Recommendations towards:
A. Use of the overall (integrated) spatial outputs produced in this project:
	Spatial outputs to be read in their entirety, i.e. including all lines of evidence.
	Spatial outputs to be used in combination with results from previous mapping studies (as additional line of evidence).
	Ground truthing recommended where spatial outputs are used for project-level assessments.
B. Improvement of the overall (integrated) spatial outputs produced in this project:
	Further stakeholder validation (esp. for maps not validated, but also for the others as validation was limited by time available in the project).
	Further collection of additional survey data for map validation, including more years and wider areas, and especially for species for which no survey data were available in the project.
	Expand habitat proxy mapping to Scotland/UK-wide inshore areas (as multiple small scale maps, for better visibility of inshore habitat proxies).
C. Improvement of EFH models and their prediction maps:
	Expand survey database on which models are calibrated to
	(i) Improve predictive ability and overall confidence of EFH models:
	- Lesser sandeel (A. marinus): include survey data with wider spatial coverage (e.g. offshore, UK-wide).
	- Nephrops: include data from Nephrops TV surveys (burrow density).
	- Anglerfish (juveniles): explore MS scallop dredge surveys and IBTS surveys for additional data inclusion.
	- All models: expanding time series considered, hence increasing dataset size and environmental range covered can improve confidence.
	(ii) Extend geographical/species coverage of EFH models:
	- Offshore/Scallops, surf clam, and possibly thornback ray juveniles, dog cockle and common whelk: collate and explore MS scallop dredge surveys for additional EFH models.
	- Offshore/Lobster and crabs: explore accessibility of offshore wind industry data sources on crustacean monitoring.
	- Inshore/shellfish (lobster, crabs, razor clam, cockles, whelks) and fish (juvenile flatfish, gadoids, sprat, skates, rays, spurdog): assess accessibility, collate and explore inshore survey data (e.g. UHI Shellfish Stock Assessment Programme, SEPA WFD fish monitoring for transitional and coastal waters) to calibrate additional models for inshore EFH habitats.
	- Inshore/environmental data layers: fill gaps in spatial coverage of environmental data layers in more internal/inshore areas through mapping from local environmental monitoring, where available (to be assessed on a case-by-case basis).
	Account for stock differences when defining aggregations to be modelled rather than doing it at species level (across different stocks) (e.g. Nephrops).
	Apply EFH model to environmental data layers for individual years and then combine outputs
	Revise EFH models periodically by including new survey data as they become available (e.g. from ongoing sandeel dredge surveys, bottom trawl surveys, Nephrops TV surveys, scallop dredge surveys).
	Explore implications of climate change on EFH distribution more accurately (identify and collate climate change spatial projections for multiple environmental model predictors, where available, and apply to EFH models with appropriate baseline and with variable time-span).
D. Improvement of habitat proxy assessment and the resulting maps:
	Extend matrix assessment to scoring of deeper habitats that were excluded a priori (because described as offshore) but that occur in inshore areas to fill spatial gaps in the maps.
	Further research on species/life stages for which information on habitat preferences was not sufficient for the assessment or for which detail was low.
	Periodically revise habitat proxy assessment matrix and update scoring and confidence as new/more information becomes available on habitat preferences of the species/life stages considered.
	Periodically update EUNIS habitat data layers used for habitat proxy mapping. Explore availability of EUNIS habitat map predictions under climate change, as these could be used to assess implications of climate change on geographical distribution of habitat proxies for species/life stages.

Recommendations towards the correct use of the overall (integrated) spatial outputs produced in this project:

• The spatial outputs of this study should be used in their entirety, i.e. considering all the lines of evidence shown, from EFH model results (including confidence), to survey data, to stakeholder feedback.
• It is also recommended that, where available, additional results from previous mapping studies are also considered as further lines of evidence to be combined with those from the present study. This would further increase the confidence on the conclusions drawn cumulatively from all the available spatial products, particularly where different lines of evidence (often based on different approaches and/or data) converge in indicating certain areas as key habitats for a species.
• The EFH models and habitat proxy assessments can be used in project-level assessments to provide an indication of areas of possible concern, i.e. areas where key habitats potentially suitable for fish and shellfish species may occur and overlap with the proposed marine project development (and its impact areas). The habitat suitability for a species or its life stage(s) may be affected by multiple abiotic and biotic variables, but some of these (or their interactions) might not be included in the EFH models or the way EUNIS habitat types are defined and identified. This may affect the accuracy of the model or habitat proxy predictions and the way these reflect the actual use of an area. Therefore, it is recommended that, where the spatial outputs of this project are to be used for project-level assessments, ground truthing surveys for the specific area of interest are always undertaken to verify the actual use by fish or shellfish at the time of the project-level assessment. Such surveys should be undertaken using the most appropriate method and design (e.g. relevant seasonality) to represent the species or life stage of concern. They could be periodically repeated to confirm possible changes in habitat use with changing environmental conditions, particularly where predictions under scenarios of environmental (or habitat) change have highlighted changes in potential suitability of the environment/habitat for the species/life stage.

Recommendations towards the improvement of the overall (integrated) spatial outputs produced in this project:

• Further map validation is recommended, especially for those maps that received no validation at all (either through survey data or stakeholder feedback), such as habitat proxy maps for European lobster (juvenile), brown crab (juvenile), velvet crab, common and dog cockles, common and dog whelk, and razor clam. However, even where map validation was possible, this was partial in some cases (based on either additional survey data or stakeholder feedback), and, in any case, it was limited by the time constraints of the project, and therefore all maps would benefit from further validation. This would entail:
  • Further stakeholder consultation on the EFH model and habitat proxy maps, with consequent inclusion of its results as one of the lines of evidence shown in the spatial products of this study. Maps for which no stakeholder feedback was received should be prioritised (inshore habitat proxy maps for small sandeel (A. tobianus), juveniles of cod, whiting, saithe, plaice, sprat and spurdog, and spawning adult/eggs of spotted and thornback ray, as well as habitats for the shellfish species mentioned above), but all maps would benefit from receiving more stakeholder feedback. A further stage of engagement with stakeholders and further validation of the maps based on the experience of the fishermen and of other fishery experts was highlighted by the Scottish Fishermen's Federation (SFF) as a key step to finalise the spatial outputs for use in environmental assessments.
  • Further data sourcing and collation to complement stakeholder input in map validation. Additional survey data could not be obtained for validation of maps for several species, including: all the shellfish species assessed in habitat proxy maps (as mentioned above); spawning habitats of cod, whiting, haddock, mackerel (eggs) (EFH model maps), and herring and Norway pout (habitat proxy maps); and juvenile common sole (both EFH model and habitat proxy maps). Data for some of these species may become available with time (e.g. current/forthcoming data from ongoing survey programmes such as IBTS for spawning gadoids, MEGS for mackerel eggs, BTS for juvenile sole), or with further exploration of existing data sources (e.g. for inshore fish and shellfish surveys; see Table 33). However, additional monitoring effort might be needed to complement map validation for species/life stages and/or areas where actual data gaps occur (to be identified based on a more comprehensive data exploration than was allowed in this project).
• The spatial implementation of the habitat proxy assessment should be extended to all areas of Scottish (and possibly UK) waters, as it was done for the mapping of herring spawning grounds, to better represent the potential distribution of essential habitats

for other species (e.g. areas surrounding Orkney for for shellfish). It is recommended that a combination of finer scale maps for different coastal areas is used, as opposed to a single map for the whole extent of Scottish inshore waters, for better visibility of smaller areas of the suitable functional habitat represented by habitat proxies. Further collection of inshore survey data from these additional areas (e.g. survey data from the Shellfish Stock Assessment Programme around Shetland) would allow the validation of these maps.

Recommendations towards the improvement of EFH models and their prediction maps:

• Data sourcing and collation may be a time-consuming task, and it was limited by the relatively short timescale of this project. The EFH model assessments undertaken in this study would benefit from extending the database used for both model calibration, particularly for those species that are of higher interest or that had poorer assessments (with lower confidence or which could not be fully assessed or validated). More specific recommendations include:
  • Catch data from sandeel surveys more widely distributed around the UK should be obtained and included in the modelling, with surveys also covering areas further offshore than those considered in the present study. This would allow a wider spatial applicability of the model (due to the wider range of environmental conditions for which the model provides valid predictions) and a likely increase in the overall confidence in the model (which was lowered for the model output developed in this project given the reduced spatial coverage of the sandeel dredge surveys used for model calibration).
  • Aggregations of Nephrops burrows from MS TV burrow density surveys available to date can be included to improve the EFH model for the species. These surveys, which directly target the refuge habitat resource (burrows), held a higher confidence compared to the bottom trawl survey data used in this project (targeting Nephrops individuals instead). As a result, an increase in the overall confidence in the model output is expected from their inclusion in the analysis.
  • Longer-term data bases could be obtained for species which had limited data in the selected study period 2010-20 to allow modelling of species/life stages that could not be covered in this study (e.g. cod juveniles) or improve the EFH model performance for species/life stages modelled with lower confidence (e.g. whiting life stages). For bottom trawl surveys (used for calibration of most EFH models), these data are available from the existing online ICES database (Datras).

More time and effort should be dedicated to identify and collate data from (i) shellfish monitoring surveys and (ii) fish monitoring inshore (e.g. fish count data obtained to comply with the Water Framework Directive), in Scottish waters as a minimum, but preferably UK- wide where possible. Potential data sources and the species that would benefit from the modelling of these additional data are as indicated in

• Table 31, and include species with EFH both offshore (e.g. scallops, surf clam) and inshore (e.g. European lobster and crab nursery, cockles and whelks, nursery grounds for flatfish, gadoids, skates and rays). It is acknowledged that, this would likely require gathering many small-scale datasets from multiple sources, and accessibility for some of these data (e.g. from the offshore wind industry sector, or WFD fish monitoring from SEPA) needs to be ascertained. Furthermore, this would require considerable additional effort to combine and standardise the different datasets into a wider-scale database that could be used to complete the modelling for fish and shellfish species and/or inshore habitats (or at least to validate the habitat proxy maps obtained based on expert assessment and literature review).
• The ability to calibrate EFH models for species in inshore areas would also depend on the availability of environmental data layers to be used as model predictors in those areas. Coverage of these data layers may not be complete inshore, particularly in most internal areas (e.g. sea lochs, estuaries), as broad-scale environmental layers are often the result of oceanographic models that have limited applicability closer to the water/land boundary (e.g. for sea temperature, water column mixing). Alternative local maps covering site-specific spatial gaps should be sourced (or data from existing local environmental monitoring should be mapped), where available, although temporal consistency with correspondent fish/shellfish survey data will need to be taken into consideration for them to be used for model calibration.
• Stakeholder feedback has highlighted that, as different stocks of a species can be characterised by different densities, the assessment on a stock-by-stock basis (i.e. identifying aggregations based on abundance within stocks rather than between stocks) is recommended to improve the model predictions (e.g. Nephrops).
• It is recommended that spatial outputs from EFH model predictions are generated for individual years (using the appropriate environmental data layers) and the results combined across years (as explained in section 4.4) to be able to account for temporal variability in the habitat distribution and persistence of use of certain areas.
• The EFH models could be periodically revised through inclusion of new survey data as they become available to strengthen the statistical tool and improve its confidence.
• A preliminary overview of the potential implications of climate change for mapped distributions of EFH was undertaken in this project (see section 3.2). However, further work should be undertaken to explore such implications more accurately. For the EFH model maps, this would mean to source (or produce, where not available) climate change spatial projections for as many of the environmental predictors used by the models (the analysis in this project was limited to sea bottom temperature and depth variability in inshore areas as a consequence of sea level rise), where possible, and with variable time span to assess possible variability on different timescales (e.g. also relevant to the life span of a project that is being assessed, from construction to decommissioning). The models could then be applied to environmental scenarios accounting for these changes (using the appropriate baseline) to assess the spatial variation in the mapped habitat distributions.

Recommendations towards the improvement of the habitat proxy assessment and the resulting maps:

• A trade-off between spatial/habitat resolution and coverage was apparent for the habitat proxy maps. Considering that the main aim of the spatial products developed in this study is to support marine spatial planning, high spatial resolution is essential. Therefore, it is recommended, that this aspect is prioritised (hence mapping habitat proxies is undertaken at EUNIS Level 4 wherever possible, using EUNIS Level 3 classification only when Level 4 is not available), even if it leads to spatial gaps in the map (where EUNIS habitat is not identified at Level 3 or 4).
• In order to reduce the spatial gaps in the habitat proxy maps, it is further recommended that the habitat proxy assessment (matrix) is extended to include those deeper habitats that were excluded a priori, but that may also occur in inshore areas.
• Further targeted research on certain species may be required so that knowledge gaps species can be filled for these species, their habitat requirements and life cycle, which prevented further assessment in this study. This should focus in particular on ascertaining habitat characteristics for juvenile and spawning individuals of sandy ray and common skate, which could not be assessed in this study. Further research into the detailed habitat characterisation for spawning thornback ray and spurdog, and juveniles of sprat, common sole, spotted ray, saithe, cod, whiting, European lobster, brown and velvet crab would also allow to increase the confidence in the habitat proxy assessments for these species/life stages. It is acknowledged that the literature review undertaken in this study was not exhaustive, hence further literature search for these species should be undertaken beforehand, to ascertain whether knowledge gaps in this study correspond effectively to gaps in the wider knowledge of these species.
• The habitat proxy assessment matrix should be maintained as a live tool, with periodic revision undertaken through expert input to allow inclusion of new/updated knowledge on the species and life stages of interest and it becomes available. This could allow expanding the range of habitat types that can be scored for a species, improving the confidence associated with the habitats already scored, or extending the assessment to other species that could not be assessed due to lack of evidence/knowledge at the time of this study.
• The EUNIS habitat data used as base layers for the spatial application of the matrix assessment should also be periodically revised, by using the most recent, high resolution habitat survey data where possible. This would allow habitat proxy maps to take into account possible changes in the underlying habitat types and their spatial distribution (e.g. due to local losses, or depth or energy changes that might alter the classification of an area into a specific habitat class).

• If predictions existed (or were developed) of possible changes in the spatial classification of areas into EUNIS habitat types (e.g. based on broadscale models accounting for future climate effects), these could be considered for the application of the habitat proxy assessment to assess resulting changes in the habitat distribution for specific species/life stages, in an analogous way as suggested for model maps (but using changes in habitat type layers in this case rather than in environmental data layers).